Direct 4D PET reconstruction with discrete tissue types.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Biophysics

Background:

• Dynamic positron emission tomography (dPET) provides spatiotemporal radio tracer data in living tissues.
• Voxel-wise kinetic modeling in dPET is often hampered by high noise levels.
• Current methods may struggle to accurately segment functional regions within the imaged volume.

Purpose of the Study:

• To present a novel direct reconstruction framework for dPET incorporating concurrent clustering.
• To address and mitigate noise issues prevalent in voxel-wise kinetic modeling.
• To develop a method that simultaneously estimates kinetic parameters, activity images, and functional clusters.

Main Methods:

• Utilized Probabilistic Graphical Modeling (PGM) theory to formalize the reconstruction problem.
• Developed an iterative estimation scheme for concurrent parameter, activity, and cluster mapping.
• Assumed the imaged volume comprises a finite number of distinct functional regions.

Main Results:

• The proposed framework successfully performed concurrent estimation of kinetic parameter maps, activity images, and segmented clusters.
• Simulation studies demonstrated the efficacy of the novel reconstruction approach.
• Exploratory application to real dPET data showed promising validation of the method.

Conclusions:

• The novel dPET reconstruction framework with concurrent clustering effectively handles noise.
• This integrated approach enhances the accuracy of kinetic modeling and functional region identification.
• The method holds potential for improved analysis of dynamic PET data in various biological and clinical applications.